Neural network–based pore flow field prediction in porous media using super resolution
Xu‐Hui Zhou, James McClure, Cheng Chen, Heng Xiao
Abstract
Predicting the pore flow velocity directly from the sub-sampled pore structure is an ill-conditioned problem. Inspired by multi-grid methods for solving systems of linear equations, we use velocity fields simulated on coarse meshes to remedy such ill-conditioning. This leads to a super-resolution-assisted geometry-to-velocity mapping for porous media.
Topics & Concepts
Porous mediumPolygon meshPorosityVector fieldGridArtificial neural networkFlow (mathematics)Flow velocityResolution (logic)Materials scienceMechanicsField (mathematics)Computer scienceGeometryMathematicsPhysicsArtificial intelligenceComposite materialPure mathematicsSeismic Imaging and Inversion TechniquesEnhanced Oil Recovery TechniquesHydraulic Fracturing and Reservoir Analysis